Search

A geometric optimization based approach to deploy a mobile sensor network for the purpose of detecting and capturing mobile targets in the plane is presented in [1]. The sensing-pursuit problem is motivated by the Marco polo game, in which the pursuer Marco must capture multiple mobile targets that are sensed intermittently, and with very limited information. In this paper we extend the results in [1] by providing (i) a complexity analysis of the proposed cell decomposition planning algorithm, and (ii) a testbed that allows to experimentally verify the applicability of the proposed pursuit methodology.

Control and optimization of acoustic sensors can significantly impact the effectiveness of sonar deployment in variable and uncertain underwater environments. On the other hand, the design of optimal control systems requires tractable models of system dynamics, which in this case include acoustic-wave propagation phenomena. High-fidelity acoustic models that capture the influence of environmental conditions on wave propagation involve partial differential equations (PDEs), and are computationally intensive. Also, by relying on the numerical solution of PDEs for given boundary and initial conditions, they do not provide closed-form functional forms for the propagation loss or other output variables. In this paper, a simple Bayesian network (BN) model of acoustic propagation is presented for use in sonar control. The performance of the BN model compares favorably to that of a radial basis function neural network. Additionally, the sensor range dependency on spatial and temporal coordinates can be estimated and utilized to compute optimal sonar control strategies.

The paper presents a three-layer feedforward backpropagation (BP) artificial neural network (ANN), whose output is battery state-of-charge (SOC), to estimate and predict SOC of high power Ni-MH rechargeable battery. Five ANN inputs are novelly selected to improve the accuracy of ANN prediction by the proposed method of correlation coefficient ranking based on correlation analysis of different variables and SOC, and they are: battery discharging current i, accumulated ampere hours Ah, battery terminal voltage v, time-average terminal voltage tav and twice time-average voltage ttav (i.e. time-average of tav). Meanwhile, six training sets are equally selected from thirteen data sets about constant current discharging (CCD) from 100 % to 0 % SOC and Levenberg-Marquardt training algorithm is selected. Comparisons between simulation and measurement verify the proposed ANN model. Especially, the ANN can satisfyingly estimate SOC of battery (pack) whose starting SOC (i.e. SOC/sub 0/) is not originally known after about ten minutes (short time compared with the whole discharging process) constant load discharging (CLD), and most of absolute values of absolute errors are not more than 5 %.

A battery is a quite complex and nonlinear system comprising interacting physical and chemical processes although it seems deceptively simple. State-of-charge (SOC), a parameter to describe how much energy battery has, is a key factor in battery management and its estimation is an important and challenging task. We develop an adaptive neuro-fuzzy inference system (ANFIS) to achieve the goal. First in this paper, nonconventional input variables of the ANFIS are selected by three different correlation analysis techniques, linear correlation analysis (LCA), nonparametric correlation analysis (NCA) and partial correlation analysis (PCA). Next, the ANFIS model of five inputs and one output is presented. Takagi and Sugeno's fuzzy if-then rules are used. Then, number determination of training data pairs is discussed. Finally, hybrid learning algorithm combining the gradient method and the least squares estimate (LSE) is adopted to train the ANFIS. Predicted results obtained by the ANFIS are compared with measured results, verifying presented ANFIS. For contrast, a three-layer feedforward back-propagation (BP) artificial neural network (ANN) is presented to estimate SOC. Compared with the BP ANN model, the ANFIS obtains better prediction performance when interpolating. Comparisons of the two approaches have highlighted the potential of ANFIS in modeling and prediction of the behavior of complex nonlinear dynamic systems.

Mobile robot often relies on a battery system as its power supply and such kind of mobile robot is called cableless mobile robot. In the past years, while there are many researches on automation and control techniques, mechanical and sensor designs of mobile robot, very little systematic and comprehensive work has been done in the design of rechargeable battery power system for cableless mobile robot. This paper discusses the above problem from three different aspects: selection of rechargeable battery type, battery management technique and energy renewal. In the section on selection of rechargeable battery, many different types of traction rechargeable battery are introduced and the factors that are crucial for cableless mobile robot applications are compared. The section on battery management technique focuses on battery state-of-charge (SOC) estimation. In the section on energy renewal, two typical charging techniques, conductive charging and inductive charging are discussed. In the final section, one implementation example, high power Ni-MH rechargeable battery system for TH-1 humanoid robot which is being developed by Tsinghua University, is introduced.

Battery is a quite complex and nonlinear system comprised of interacting physical and chemical processes, although it may seem deceptively simple. For this reason, existing battery models are either partially successful or too complicated and inconvenient in modeling and simulation applications. This paper presents a dynamic neural network model with time-delayed system output feedback for modelling and identification of a Ni-MH battery during discharging. Comparisons between the simulation and measurement verify the presented model. Compared with the methods based on the Peukert equation, which is often used for the calculation of the available capacity and simulation of battery discharging curves, the ANN method is more accurate.